Background

In December 2019, SARS-CoV-2 unprecedentedly spread around the world, overwhelming global healthcare systems. On March 11, 2020, the World Health Organization declared the coronavirus disease 2019 (COVID-19) global pandemic. This led to a rippling impact of the virus on healthcare systems. In order to reduce viral transmission and relieve pressure on healthcare networks, many countries, including the United Kingdom (UK), entered a national lockdown by which people were ordered by law to stay at home [1. ]. In many hospitals, staff were redeployed, and departments were adapted or converted to provide COVID-19 services [1. ].

Whilst public health emergencies explicitly effect the physical health of a population, increased levels of poor mental health can also be discovered (e.g., depression, PTSD, substance use disorder, and behavioural disorders) [2. ]. Influences directly related to infection, such as, the neuroinvasive potential of SARS-CoV-2, may affect brain function and in turn mental health. The treatment for COVID-19 may also have adverse effects on mental health. For example, the imposition of unfamiliar public health measures (i.e., social isolation) increases the likelihood of clinically significant depression or anxiety [2. , 3. ]. Whilst all individuals were urged to comply with lockdown protocols, emotional distress tempted some to consider violating the recommended public health measures [3. ].

One vulnerable group during the pandemic were pregnant women and women who had recently given birth. Millions of women experience mental ill-health during pregnancy and after childbirth, with maternal mental ill-health being an international public health concern, affecting up to 10% of women during pregnancy and 13% of women after childbirth [4,5,6. ]. Compromised mental health can cause short and long-term consequences for the mother and baby however limited data exists on the prevalence of mental ill-health in women who were pregnant and gave birth during the COVID-19 pandemic [6,7,8. ].

This systematic review and meta-analysis will assess the prevalence of mental ill-health in women during pregnancy and after childbirth throughout the Covid-19 pandemic. Findings with be compared to other global pandemics including SARS and MERS.

Methods

A systematic methodology was developed along with a relevant protocol that was peer reviewed and published in PROSPERO (CRD42021235356).

Search criteria

The search criteria were developed based upon the research question using PubMed, Science Direct, Ovid PsycINFO and EMBASE databases. A wide search criterion was used to ensure the inclusion of all pregnant women with existing gynaecological conditions. The MeSH terms used include (COVID) OR (SARS-CoV-2) AND (SARS) AND (MERS) AND ((mental health) OR (depression) OR (anxiety) OR (PTSD) OR (psychosis) OR (unipolar) OR (bipolar)) AND ((PCOS) OR (fibroid) OR (endometriosis) OR (pre-eclampsia) OR (still birth) OR (GDM) OR (preterm birth) OR (women’s health) OR (pregnant women) OR (pregnancy)).

Screening eligibility criteria

All studies published in English were included from 20th December 2019 to 31st July 2021. Screening and data extraction were performed by two authors independently. Initially, titles and abstracts were reviewed to determine the relevance. A PRISMA (Preferred Reporting Items for Systematic Reviews and Meta-analyses) diagram was completed based on the eligibility steps completed (see Fig. 1).

Fig. 1
figure 1

PRISMA Flow Chart outlining search strategy. Legend. The above PRISMA flow chart outlines the study search strategy for both the systematic review and meta-analysis

Full size image

Inclusion and exclusion criteria

All COVID-19, SARS and MERS studies that evaluated the mental health of pregnant women with/without gynaecological conditions that were reported in English between December 2000 – July 2021 were included. All other studies were excluded from this analysis.

Data extraction

Full texts of the included papers were reviewed to extract the following data: time and locations of the study, participants demographics, sample size, mean age, gestation, days since childbirth, prevalence of mental symptoms, data collection tools used, and cut-offs scores applied. Any disagreements were discussed and resolved by consensus between two authors. For studies with both COVID-19 and non-COVID-19 cohort, we only used data of the COVID-19 cohort and the p-value comparing them. Studies from SARS and MERS were also reviewed in full to ensure the eligibility criteria was met. Studies reporting mean (SD) or median (IQR) of the scales measuring mental symptoms instead of prevalence rates were included and a simulation method assuming normal distribution was applied to generate the corresponding prevalence rates.

Risk of bias (RoB) assessment

A risk of bias (RoB) table assessment was completed to demonstrate the risk of bias within the studies used in the systematic review and meta-analysis. The RoB is reflective of a fixed set of biases within domains of study design, conduct and reporting. This combined with the quality check allows the findings of the study to be scientifically justified, and clinically viable. The Newcastle-Ottawa-Scale (NOS) was used to assess the RoB for all systematically included studies as demonstrated within the RoB table (See Page 33, Table 1).

Table 1 Risk of Bias for all included studies
Full size table

Data analysis

Random effects model with restricted maximum-likelihood estimation method was applied for the meta-analysis and I-square statistic was used to evaluate heterogeneity across studies. The pooled prevalence rates of symptoms of anxiety, depression, PTSD, stress, and sleep disorders with 95% confidence interval (CI) were computed. Subgroup analysis was conducted based on pregnancy trimester. Sensitivity analysis was performed to assess the robustness of the results. Potential publication bias was assessed with funnel plot and Egger’s test. Analyses were conducted with the R studio (version 1.4.17.17) and STATA 16.1.

Results

Our initial search identified a total of 1603 papers and 523 studies were excluded after screening by titles and abstracts. After full-text evaluation, 217 were included in the systematic review and 99 studies were included in the meta-analysis. The PRISMA flowchart was illustrated in Fig. 1.

Characteristics of studies

A total of 217 COVID-19 studies were included and 99 studies were meta-analysed. These studies were reported from various parts of the world, as indicated in the characteristics (See page 35 for Tables 2 and 3). We did not identify SARS and MERS studies that were suitably aligned to the eligibility criteria of our study.

Table 2 Outline of all studies in the systematic review and meta-analysis
Full size table
Table 3 Studies selected for the meta-analysis
Full size table

Study design, source of data, data collection method and sample size

All 217 studies used different study designs; 107 cross-sectional, 7 cohort and 7 case controlled. A total of 23 qualitative studies used self-reported methods of data collection. Real-world data from hospital admissions were used in 5 studies whilst 2 extracted data from patient medical records. The 217 study-pool comprised of a sample of 638,889 pregnant women, including 6898 women who were within 90 days of delivery. The sample sizes used within the studies varied considerably; 129 studies comprised of approximately 500 women, 40 studies consisted of500–999 ladies, 18 studies had 1000–1999 women and 24 studies had ≥2000 women.

Stages of pregnancy assessed

A total of 99 studies reported pregnant women during their first, second and third trimester.

Site of data collection

Many studies reported that data collection took place during routine antenatal or postnatal visits in outpatient departments, tertiary/provincial hospitals, secondary level or district hospitals, and primary healthcare facility level.

Mental health outcomes assessed

Sixty-four reported data on depressive symptoms, 82 on symptoms of anxiety, 20 on symptoms of stress, 7 on PTSD symptoms, and 8 on symptoms of sleep disorders. Detailed characteristics of the systematically included studies and those meta-analysed are listed in Tables 1 and 2 (see page 33 for Table 1 and page 35 for Table 2).

Meta-analysis

Depression

Edinburgh Postnatal Depression Scale (EPDS), the Patient Health Questionnaire 9-item (PHQ-9), and the depression subscale of the Hospital Anxiety and Depression Scale (HADS-D) were the commonly used data collection tools to assess symptoms of depression. The pooled prevalence of depression was 24.91% with a 95% CI of 21.37–29.02% (see Fig. 2).

Fig. 2
figure 2

Forest plot showing prevalence of depressive symptoms. Legend. The forest plot shows the first author, sample size and geographic location for the included studies. Measure of depressive symptoms, prevalence rate with confidence internal and weighting of results have also been displayed

Full size image

Anxiety

Anxiety symptoms were commonly measured by the State-Trait Anxiety Inventory (STAI, with two subscales STAI-T and STAI-S), the General Anxiety Disorder 7-item (GAD-7), and Self-rating Anxiety Scale (SAS). Anxiety prevalence was 32.88% with a 95% CI of 29.05 to 37.21% (see Fig. 3).

Fig. 3
figure 3

Forest plot showing prevalence of anxiety symptoms. Legend. The above forest plot shows the first author, sample size and geographic location for the included studies. Measure of anxiety symptoms, prevalence rate with confidence internal and weighting of results have also been displayed

Full size image

Str ess

Tools like the Perceived Stress Scale (PSS, with 10-item and 14-item versions), and the stress subscale of the 21-item Depression Anxiety and Stress Scale (DASS21-S) were frequently used to evaluate stress symptoms. The pooled prevalence of stress among perinatal women was 29.44% (95% CI: 18.21–47.61%) (see Fig. 4).

Figure 4
figure 4

Forest plot showing prevalence of stress symptoms. Legend. The above forest plot shows the first author, sample size and geographic location for the included studies. Measure of stress symptoms, prevalence rate with confidence internal and weighting of results have also been displayed

Full size image

Post-traumatic stress disorder

PTSD symptoms were typically measured by the DSM-V Post-Traumatic Stress Disorder Checklist (PCL-5) and the Impact of Events Scale (IES). The studies reporting PTSD symptoms were heterogeneous resulting in a pooled prevalence of 27.93% with a 95%CI of 9.05–86.15% (see Fig. 5).

Fig. 5
figure 5

Forest plot showing symptoms of PTSD. Legend. The above forest plot shows the first author, sample size and geographic location for the included studies. Measure of PTSD symptoms, prevalence rate with confidence internal and weighting of results have also been displayed

Full size image

Insomnia

The Insomnia Severity Index (ISI) and the Pittsburgh Sleep Quality Index (PSQI) were to assess and report symptoms associated with sleep disorders. The pooled prevalence was 24.38% with a 95% CI of 11.89–49.96% (see Fig. 6).

Fig. 6
figure 6

Forest plot showing symptoms of sleep disorders. Legend. The above forest plot shows the first author, sample size and geographic location for the included studies. Measure of sleep disorder symptoms, prevalence rate with confidence internal and weighting of results have also been displayed

Full size image

Subgroup analysis

The I2 evaluated for symptoms of depression, anxiety, PTSD, stress, and sleep disorders were over 98%, which demonstrates a high heterogeneity among the studies. Therefore, a subgroup analysis was conducted to further evaluate the heterogeneity. To determine the symptom prevalence, women were assessed at different stages of their pregnancy and the dataset was categorised based on the trimesters:1st trimester (< 12 weeks), 2nd trimester (13–27 weeks), 3rd trimester (28–41 weeks)] and the immediate post-partum period (immediately after childbirth and up to 6 weeks) for studies that reported follow-up details.

The heterogeneity of depressive symptoms was lower in comparison to anxiety, PTSD, stress, and sleep problems. Heterogeneity within the 1st trimester was 89.47%. I2 of the anxiety group during the 1st trimester and 2nd trimester were 88.91 and 92.35%, respectively. These appear to be similar to the I2 values of depression. I2 for stress associated with the 2nd and 3rd trimesters were 78.57 and 64.65%, respectively, indicating mild heterogeneity. Intuitively, Maharlouei and colleagues study reported a small prevalence, thus could be an influencing factor for the heterogeneity reported. I2 for PTSD across three trimesters were 24.67, 89.47 and 81.62%, respectively. I2 was 0% during the 1st trimester within the groups of participants reporting sleep disturbance. 1st trimester group showed relatively low heterogeneity across mental health symptoms, thus strictly stipulating the gestational weeks of the included pregnancy helped reduce the heterogeneity. Forest plots were generated for 1st trimester, 2nd trimester, 3rd trimester, post-partum and overall, for symptoms of depression (see Figs. 7, 8, 9, 10 and 11), anxiety (see Figs. 12, 13, 14, 15 and 16), stress (see Fig. 17), PTSD (see Fig. 18), sleep disorders (see Fig. 19). Funnel plots were also generated: depression (see Fig. 20), anxiety (see Fig. 21), stress, (see Fig. 22), PTSD (see Fig. 23), and sleep disorders (see Fig. 24).

Fig. 7
figure 7

A forest plot showing the subgroup analysis for depressive symptoms in the 1st trimester. Legend. First author and outcomes measures have been included. Prevalence rate with confidence internal and weighting of results have been shown

Full size image
Fig. 8
figure 8

A forest plot showing the subgroup analysis for depressive symptoms in the 2nd trimester. Legend. First author and outcomes measures have been included. Prevalence rate with confidence internal and weighting of results have been shown

Full size image
Fig. 9
figure 9

A forest plot showing the subgroup analysis for depressive symptoms in the 3rd trimester. Legend. First author and outcomes measures have been included. Prevalence rate with confidence internal and weighting of results have been shown

Full size image
Fig. 10
figure 10

A forest plot showing the subgroup analysis for depressive symptoms postpartum. Legend. First author and outcomes measures have been included. Prevalence rate with confidence internal and weighting of results have been shown

Full size image
Fig. 11
figure 11

A forest plot showing the overall subgroup analysis for depressive symptoms. Legend. First author and outcomes measures have been included. Prevalence rate with confidence internal and weighting of results have been shown

Full size image
Fig. 12
figure 12

A forest plot showing the subgroup analysis for anxiety symptoms in the 1st trimester. Legend First author and outcomes measures have been included. Prevalence rate with confidence internal and weighting of results have been shown

Full size image
Fig. 13
figure 13

A forest plot showing the subgroup analysis for anxiety symptoms in the 2nd trimester. Legend. First author and outcomes measures have been included. Prevalence rate with confidence internal and weighting of results have been shown

Full size image
Fig. 14
figure 14

A forest plot showing the subgroup analysis for anxiety symptoms in the 3rd trimester. Legend. First author and outcomes measures have been included. Prevalence rate with confidence internal and weighting of results have been shown

Full size image
Fig. 15
figure 15

A forest plot showing the subgroup analysis for anxiety symptoms postpartum. Legend. First author and outcomes measures have been included. Prevalence rate with confidence internal and weighting of results have been shown

Full size image
Fig. 16
figure 16

A forest plot showing the overall subgroup analysis for anxiety symptoms. Legend First author and outcomes measures have been included. Prevalence rate with confidence internal and weighting of results have been shown

Full size image
Fig. 17
figure 17

A forest plot showing the subgroup analyses for stress symptoms. Legend. The forest plot shows the subgroup analyses results during the 1st, 2nd, and 3rd trimester as well as postpartum and overall. First author and outcomes measures have been included. Prevalence rate with confidence internal and weighting of results have been shown

Full size image
Fig. 18
figure 18

A forest plot showing the subgroup analyses for PTSD symptoms. Legend. The forest plot shows the subgroup analyses results during the 1st, 2nd, and 3rd trimester as well as postpartum and overall. First author and outcomes measures have been included. Prevalence rate with confidence internal and weighting of results have been shown

Full size image
Fig. 19
figure 19

A forest plot showing the subgroup analyses for sleep disorder symptoms. Legend. The forest plot shows the subgroup analyses results for sleep disorder symptoms during the 1st, 2nd, and 3rd trimester as well as postpartum and overall. First author and outcomes measures have been included. Prevalence rate with confidence internal and weighting of results have been shown

Full size image
Fig. 20
figure 20

Funnel plot of depressive symptoms

Full size image
Fig. 21
figure 21

Funnel plot of anxiety symptoms

Full size image
Fig. 22
figure 22

Funnel plot of stress symptoms

Full size image
Fig. 23
figure 23

Funnel plot of PTSD symptoms

Full size image
Fig. 24
figure 24

Funnel plot of sleep disorders symptoms

Full size image

Publication bias and sensitivity analysis

Publication bias and sensitivity analysis tests were conducted to assess the reliability of the data as some studies had large standard errors that would produce undesirable effects. Copas Selection Model was used to select studies for the sensitivity analysis. The p-values of residual selection bias were evaluated (see Fig. 25, 26, 27, 28 and 29). Studies with a p-value of > 0.1 indicated that the residual selection had minimal bias and, the selected studies can be represented. The proportions identified were 67.84, 100 and 59.49% for depression, anxiety, and sleep disorders, respectively. For studies reporting stress and PTSD, the Copas Selection Model could not provide a decision, indicating the previous conclusions of high heterogeneity is accurate.

Fig. 25
figure 25

P-value for residual selection bias of depressive symptoms

Full size image
Fig. 26
figure 26

P-value for residual selection bias of anxiety symptoms

Full size image
Fig. 27
figure 27

P-value for residual selection bias of stress symptoms

Full size image
Fig. 28
figure 28

P-value for residual selection bias of PTSD symptoms

Full size image

A summary of studies used within the Copas Selection Model and Random Effects Model indicate that the two models have no significant difference (see Table 4). P-value of the changes between these conclusions were 0.1108 for depressive symptoms, 0.638 for anxiety symptoms, and 0.1042 for sleep disorder symptoms. The p-value of the Egger’s test was 0.0256 for studies of depressive symptoms, revealing the existence of publication bias (see Table 5). The p-values of 0.256 and 0.998 indicate that it is challenging to detect publication bias for studies associated with symptoms of anxiety and sleep disorders (see Table 5).

Table 4 Summary of sensitivity analysis
Full size table
Fig. 29
figure 29

P-value for residual selection bias of sleep disorders symptoms

Full size image
Table 5 P-value of Egger Test for the five mental health symptoms
Full size table

Discussion

Main findings

Our study demonstrates that symptoms of depression, anxiety, PTSD, stress, and sleep problems were common throughout the pregnancy period and after childbirth during the COVID-19 pandemic with 24.9% of women reporting symptoms of depression, 32.8% anxiety, 29.44% stress, 27.93% PTSD, and 24.38% sleep disorders. The lack of research conducted to assess the mental health impact of SARS and MERS on pregnant women is a significant limitation as such data could support preparation for similar pandemics in the future. Our meta-analyses indicate the clear impact of COVID-19 on the mental health of pregnant and post-partum mothers, with a pooled prevalence of the multiple symptomatology of depression, anxiety, PTSD, stress, and sleep disorder.

Strengths and limitations

To our knowledge, this is the first systematic review and meta-analysis to focus on mental health outcomes in women during pregnancy and after childbirth during the Covid-19 pandemic. The searches were not limited by geographical location or language, therefore, further increasing the chances for all relevant literature to be identified. The MESH terms used did not consider all types of obstetric or gynaecology conditions but did include the common conditions. The variety of screening tools used across the included studies must be considered when interpreting the results of this review. Direct comparisons cannot be made where the same screening tool was not used. Furthermore, most studies used self-reported questionnaires, with no clinical follow-up to confirm diagnoses. Therefore, the results cannot be interpreted as prevalence of mental illness, but rather prevalence of symptomatology.

Interpretation

Similar to our study, other research has demonstrated that the extent and severity of mental health impacts increased in women throughout pregnancy and after childbirth during humanitarian disasters and pandemics [224. ]. The subgroup analysis showed that the prevalence of symptoms of depression ranged from 16.52 to 24.96% across the four time points. In terms of anxiety symptoms, prevalence ranged from 22.06 to 32.09%. Likewise, Grumi et al. (2021) found prevalence of depressive and anxious symptoms ranges between 26 and 32% amongst pregnant women through the COVID-19 pandemic [225. ]. Contrary to previous findings, we found that pregnant women and women who have just given birth experience higher levels of anxiety, especially in the 1st trimester and post-partum, compared to depressive symptoms [226. ]. In terms of symptoms of anxiety and PTSD, some research has found that these symptoms have been elevated in pregnant women throughout the COVID-19 pandemic [24. ]. Women who became pregnant or gave birth during the pandemic suffered from various symptoms of poor mental health across all stages of their pregnancy and postpartum. It is unclear as to the reason for this observation, and the impact of this in a real-time scenario.

These findings could be due to pressure of being a first-time mother or, general stress and health anxiety regarding how and when to access care from midwives and obstetricians as part of routine and emergency maternity care due to the Covid-19 pandemic. Similar to our findings, other studies carried out during the Covid-19 pandemic reported up to 70% of pregnant women suffering from stress during the pandemic [8. ]. Being pregnant and giving birth are known triggers for women to develop anxiety and depression and is a known risk factor for exacerbations or decline in pre-existing mental ill-health [5. , 6. ]. Other possible reasons for the increase in mental ill-health in women during pregnancy or after childbirth may be because of the massive clinical changes that took place regarding how women could access maternity care during the Covid-19 pandemic. As pregnant women were at higher risk of severe illness if infected with SARS-CoV-2, they advised to be stringent with public health measures such as social distancing and self-isolation to lower their risk of COVID-19 exposure. This led to the rapid implementation of virtual access to antenatal care in order to minimising the need for travel to antenatal clinics and in-person contact with healthcare staff. Antenatal care changed immediately from face-to-face consultations to telephone or video consultation. Birth partners were limited in number, with visiting hours for partners restricted resulting in less emotional and psychological support for women during labour and after childbirth on the postnatal wards. Furthermore, once the Covid-19 vaccination was developed, there was uncertainty regarding the effectiveness and safety of the vaccine for pregnant women, which may also have contributed to and exacerbated stress and anxiety.

Recommendations

All women should be risk assessed for maternal mental health at their initial visit with antenatal services and screened at every contact during pregnancy and after childbirth. All healthcare systems need to invest in perinatal mental health services, delivered from a multi-disciplinary team including mental health nurses, specialist midwives, obstetricians with specialist interest in mental health, and perinatal psychologists and psychiatrists. Maternity mental health services should be delivered in a way that meets the specific needs of the individual patient, including face-to-face consultations, telephone calls and/or video consultations. Up to date information regarding the impact of Covid-19 on maternity services needs to be available and accessible for women during pregnancy and after childbirth (e.g., through social media campaigns or hospital websites). Learning from this data, considerations of the special needs of the pregnant and postnatal mothers should be imperative in the implementation of strategies improve preparedness of the health service in future pandemics.

Conclusion

This study highlights that maternity mental ill-health was common during the Covid-19 pandemic and highlights the need to understand the complexity of factors associated with maternal mental health. Maternity mental health services need further investment and prioritisation with clear effective referral pathways and support for women who report mental health concerns during and after pregnancy. Further research is required to explore how to best provide care in a way that meets the specific needs of each woman, across different healthcare systems.